Zama Brings FHE Model Support to HF Endpoints

By leveraging Zama's Concrete ML framework, models with Fully Homomorphic Encryption (FHE) technology applied can be deployed on Hugging Face Endpoints. FHE allows computations to be performed on encrypted data without decrypting it, providing strong protection for user data privacy.

The key features and usage methods are as follows:

• Easy Deployment: Using Hugging Face's custom inference handlers, FHE-friendly models pre-compiled by Zama can be deployed with just a few clicks.
• Hardware Requirements: Currently, Concrete ML models run on CPU rather than GPU environments, with GPU support planned for the future.
• Client Integration: After installing Concrete ML on the client side, users can send encrypted data through the deployed endpoint API and receive inference results.